Lane Detection System Filtering False Positives
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing lane recognition systems suffer from false lane detection issues, which can lead to accidents and are not suitable for autonomous driving applications.
Innovation Solution
A lane detection system and method that extracts feature points from road images, selects candidate start and end points based on lane width and control points, and filters non-connected points using road map information to minimize false detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional lane recognition systems are used, then lane detection can be performed, but false detection occurs frequently
Solution Approach 1:
The patent segments the lane detection process into distinct stages: extracting feature points from the road image, selecting candidate control points from feature points, determining candidate start and end points from control points, and finally selecting optimal start and end points. This segmentation allows each stage to refine the results progressively, reducing false detections by eliminating unreliable points at each step rather than processing all points uniformly.
Solution Approach 2:
The patent applies local quality by using different selection criteria for different parts of the detection process. Candidate control points are selected based on local geometric relationships with feature points, candidate start/end points are determined based on local patterns in the candidate control points, and final start/end points are selected based on their positional relationships with the road and lane markings. Each local region is processed with appropriate criteria suited to its characteristics.
2Measurement precision
If more feature points are extracted to improve detection coverage, then more candidate points are available, but false detection increases
Solution Approach 1:
The patent performs preliminary filtering actions before final lane determination. Candidate control points are pre-selected from feature points based on geometric criteria, and candidate start/end points are pre-identified from control points based on pattern recognition. These preliminary selections reduce the number of points that need further processing and eliminate obviously incorrect candidates early in the process.
Solution Approach 2:
The patent implements feedback mechanisms where the selection of candidate control points depends on the distribution and characteristics of extracted feature points, and the selection of final start/end points depends on the patterns observed in candidate control points and their relationships with road geometry. This feedback loop allows the system to adapt to different road conditions and lighting scenarios, improving accuracy while maintaining reliability.
Data Source
AI summary
Provided is a lane detection system and method. The lane detection method includes extracting feature points of a driving lane from an input road image, extracting candidate start points and candidate end points from candidate control points selected from among the feature points, selecting a left start point and a right start point of the driving lane from among the candidate start points based on information corresponding to a width of the driving lane, and selecting a left end point and a right end point of the driving lane from among the candidate end points based on a number of the candidate control points between each of the left start point and the candidate end points and the right start point and the candidate end points.